Siru

This seems like the way it’s going to go. For large companies there is no doubt that LLMs can be helpful. But running your own hardware has been proven over and over to be less costly in the long term. So I can’t imagine that subscription models like Claude and ChatGPT will be the final way it works.

Just to make sure I got this right, this sounds like someone is running around with a (potentially) breached phone in the white house?

YES! An extra backup never hurts! (Well it kind of does with these storage prices at the moment, but nonetheless…)

Third party insurance often is only a minimum that you have to fulfill. E.g. IN USA has a law stating that you must have $25k liability insurance as a driver. If you cause damages beyond that then you privately are liable again. So in this case it also wouldn’t make everybody’s insurance go up.

If you were the one unlucky dude who caused a wreck with the $1 trillion car, then you would just privately be f**ked.

On a side note though: A lot of these kinds of court cases end up going for settlements much lower than the actual damages caused, because you can only recover the amount somebody actually has in court. If the damages are $1000 and the other person only has $10 to their name, you will only get $10 from them (and you will still have to pay your own layer costs from that as well). So driving this expensive car is really only dangerous to the owner who will lose a bunch of money if it gets damaged.

Maybe this is different in different countries, but I don’t think insurance has to agree to cover a given vehicle for a specific person. So basically, if you try to insure it, they could just say no.

they’ll just see it again tomorrow

This is sort of the implied contract of teaching somebody something or reviewing for them. They have to get better over time. If they do not, then I have zero interest spending my time on them. Irrelevant of whether they are a human or an AI.

Exactly, there is every intention of them coming back up again. It seems at most like their servers got hacked, and at least that they need (to schedule) maintenance to keep things running smoothly.

If they were gone forever, or there was no published information on what is going on, I imagine much fewer authors would have an issue with their creations being carried on (in good faith) on a new site.

Will Google have a reason to do that though? Won’t they just leave the contractors to fend for themselves?

If they are doing this well, why are they trying to get rid of a bunch of employees?

Has pure HTML (no css, no javascript) ever had a serious security issue? On the server side obviously not (outside of the apache/nginx/etc. having application bugs), but even on the client side has there ever been an exploit?

One of the original ideas behind Markdown was that a document should remain readable even before a computer interprets its formatting.

The author states this towards the beginning of the article. This is possible everywhere you can easily open a txt file. Note, there are quite a few platforms that don’t like to make it easily possible to open text files.

Then later, he goes on to say that opening it should interpret the formatting as well. This is neither the original premise of Markdown, nor is it easily doable. Namely, there are a number of different markdown ‘specs’: Github-flavored, Commonmark, etc. On top of that, Markdown is (mostly) a superset of html syntax. Technically any html is valid markdown formatting.

Thus, in order to ‘simply’ render the formatting of Markdown, you first need information on what type of markdown, and then also be running a fairly complex parsing+rendering pipeline.

Not to mention that you are doing all that for a document format that the non-tech/average user has never even heard of probably. This is the same reason why syntax highlighting (beyond Regex/Treesitter) isn’t part of most text editors.

supposed to …

For the record, this is just windows. MacOS and (most) flavors of Linux use the standard SI units. And also, just because computers are base2 for the mathematics inside the processor, that doesn’t mean they have problems dealing with number other than 2^n. That would be equivalent to saying that humans can only use numbers like 10, 100, 1000 and nothing in between.

(The only area where things do get difficult is calculating with fractional values, such as 1/3 because the count of decimals needed to represent can vary depending on the base. For example, 1/3 is difficult to represent in base10 because it is infinite 0.333… however in base3 it requires only a single ‘decimal’/fractional digit 0.1.)

Is there a way to access this without having to pay for/have an account?

This article reads so superficially. I would have hoped that a university would put a bit more of the actual mass of the research into their article about it.

Considering that the main argument is that supply chains are difficult, I don’t see how, for example, the US could confirm that China is no longer internally producing these AI training chips or vice versa. Aren’t the supply chains in different countries somewhat disjoint?

The problem is the spam so largely outweighs the actual hand-crafted, good ones, that just getting rid of it all is the only real solution for the end user.

\s(?) Why must these quotes always be so old already…

Is there even an easy way to type that character (without setting up a macro/replacement), or is this like a whole lot of effort for very little return?

So cool! I don’t think I’ve ever seen a braille image render before.

Asanoying as it is to change a habit, I would be reallgglad if alotof these things could become standardized accross different products. After all, I don’t speak a different variety ofEnglish just because I switched between Feams, Slack, Discord… so why should anything else I am typing i to the text fields have to change?